IFMBE Proceedings Vol. 14/1 Volume 1 Track 01 77 A model predictive control strategy for the regulation of hypnosis Yelneedi Sreenivas, S.Lakshminarayanan 1 and G.P.Rangaiah Department of Chemical and Biomolecular Engineering, National University of Singapore, Singapore Abstract— The “anesthetic state” is a dynamic combina- tion of hypnosis, amnesia, analgesia, neuromuscular and neu- rohumoral blockade. To achieve this state, different combina- tions of drug effects must be induced in the patient undergoing surgical procedure under anesthesia. Patients are routinely and extensively monitored while under anesthesia. This makes it easy to observe and quantify the effects of anesthetic drugs and also to establish the relations between the drugs adminis- tered, degree of effect and the factors that can influence this relation. This work focuses on the problem of controlling the hypnosis during an anesthesia process by automatic regulation of a hypnotic drug. We assume that the remaining anesthetic states are regulated by manual administration of specific drugs. Past research in this area has employed the cascade control strategy for the automatic control of hypnosis by using the bispectral index (BIS) as the primary controlled variable and drug concentration as the second-dary controlled variable. Note that the BIS is an indirect measure of the hypnosis. We use a nonlinear pharmacokinetic-pharmacodynamic represen- tation of the hypnosis process dynamics and propose a model predictive control (MPC) structure for controlling the BIS. This strategy has found many applications in chemical process control and has potential for medical applications. Further, the constraints imposed upon the variables and saturation limits imposed on the actuators are very well taken care of by the MPC strategy. This controller also performed well even when one of the measured feedback signals failed. The results ob- tained from the MPC strategy are compared with the other control strategies (adaptive control strategy with modeling error compensation) reported in the literature. This article gives an insight into the application of different control strate- gies to the automatic regulation of hypnosis using BIS as the controlled variable. Keywords— Control of hypnosis, Bispectral index (BIS), Model predictive control I. INTRODUCTION Hypnosis describes a state of anesthesia which is related to unconsciousness and also to the disability of the patient to recall (amnesia). The design of drug infusion policies to regulate the hypnosis is an important control problem in biomedical processes. In clinical practice, anesthetists administer drugs and adjust several infusion devices to achieve specific desired anesthetic state in the patient. In this, anesthetist is acting as a manual feedback controller. Automatic regulation of infusion of drugs during general anesthesia has several advantages when compared to man- ual administration [1]. Some of the important benefits of automatic control system for control of hypnosis is to avoid both overdosage and underdosage of the drugs, to minimize the drug consumption thereby decreasing the cost of the surgery, and also to reduce the inter-individual and intra- individual variability to compensate for differences in sur- gical procedures and anesthetic regimes. Design of a feedback controller requires a reliable mathe- matical model of the patient and appropriate monitoring devices to measure the level of hypnosis. To measure the hypnosis level, a commercial monitor (BIS index, Aspect Medical systems, Newton, MA), is available. BIS is an electroencephalogram (EEG) derived parameter which provides a continuous reliable estimate of hypnotic state [2]. The mathematical model which is developed is a series combination of pharmacokinetic (PK) model and pharma- codynamic (PD) model. The PD model has an effect com- partment which is attached to the central compartment to compensate for transport resistances. The closed loop control approach proposed by Gentilini et al. [3] used the internal model control (IMC) framework to design the controller for hypnosis. The controller design has a cascade control structure, which consists of a master controller which controls the BIS and a slave controller to regulate the endtidal concentration. The clinical results showed that the controller worked well when compared to manual administration [4]. Recently, Puebla and Alvarez [5] proposed an adaptive feedback controller for regulation of hypnosis based on modeling error compensation (MEC) ideas. They proposed that the MEC design takes care of the patient variability very well. A cascade control structure is used to design the feedback controller to regulate the hypnosis level and also it includes the MEC approach. Results showed that, the con- troller performed better than the controller design proposed by Gentilini et al. [3]. But, this work [5] is limited to only simulation studies and has not been tested clinically. The contribution of the present article is to propose a model predictive controller for automatic regulation of hypnosis level of the patients. One of the important proper- ties of MPC is the ability to handle constraints. The pro- posed MPC will regulate both BIS and endtidal concentra- tion by manipulating the infusion rate of isoflurane.